Affinity-based RL proposed for virtuous AI agent behavior

Arxiv CS.AI (top papers)

Paper introduces affinity-based reinforcement learning, using policy regularization to incentivize virtuous behavior in AI agents. The technique is demonstrated in a game environment, aiming to engineer ethical actions through reward shaping.

reinforcement-learningai-safety

Arxiv CS.AI (top papers)·Ajay VishwanathChristian Omlin

7 days ago